Eric Tang
San Francisco, California, United States
Summary
🤩
RockstarEric Tang is a software engineer focused on LLMs with six years of experience building and evaluating large-scale ML systems across industry and academia. Currently at Anyscale working on post-training systems like SkyRL, he has interned at Google DeepMind and contributed to research at Stanford SAIL, blending practical engineering with novel research. His open-source work on EleutherAI’s lm-evaluation-harness includes adding and refining math evaluation tasks, showing strength in dataset/task engineering and robust parsing for model evaluation. Past internships at Meta, TikTok, and Accenture reflect a breadth across ads optimization, computer vision, and systems for model serving. Trained at Stanford (MS CS) and UC Berkeley (BS Materials Science), he combines rigorous academic grounding with hands-on production experience in the Bay Area.
6 years of coding experience
Github Skills (6)
transformer10
evaluation-framework10
language-model10
python10
string-parsing9
testing6
Programming languages (4)
Github contributions (5)
EleutherAI/lm-evaluation-harness
Mar 2021 - Mar 2021
A framework for few-shot evaluation of language models.
Role in this project:
ML Engineer
ML EngineerContributions:7 commits, 1 PR in 5 days
Contributions summary:Eric primarily contributed to the `lm-evaluation-harness` repository by adding and modifying math tasks within the evaluation framework. Their work involved defining the structure of math problems, implementing functions to process results, and handling string parsing for answer extraction. They also made iterative updates to the code, including fixes for string parsing and removal of testing code.
language-modelevaluation-frameworktransformer
erictang000/SkyRL
Jun 2025 - Jul 2026
SkyRL-v0: Train Real-World Long-Horizon Agents via Reinforcement Learning
Contributions:2 PRs, 553 pushes, 176 branches in 1 year 1 month
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.